bxl1.hydramirror.experiencenet.com returns HTTP 503 at the edge while the backing container is healthy. Split out of issue #447 (pi-node-003 recovery, issue #445).
RESOLUTION (2026-08-06)
Hypothesis 1 confirmed, with a nuance: the stale record was not an inventory-refresh problem but a STATIC route in hydrascalerouter config on the edge (141.227.136.199), /root/.hydrascalerouter/config.yaml. The 2026-08-04 renumbering edit (see config.yaml.bak-20260804-160721) updated hydrabody (10.0.0.5 -> 10.10.100.18) and added bxl1-b, but left bxl1.hydramirror at pre-renumbering 10.0.0.4:30003. Static routes carry Node=config/Scale=static in routes.json and are never refreshed by node inventory reports, which is why node-side rebind, hydraskin-report restart and the pi-node-003 reboot could not help.
Fix: updated the static route backend to http://10.10.100.19:30003 (verified 200 from the edge over the mesh; 10.0.0.4 unreachable), backup at config.yaml.bak-20260806-issue448, restarted hydrascalerouter. Verified: https://bxl1.hydramirror.experiencenet.com/api/v1/health -> 200, / -> 404 (normal mirror pass-through, same as bxl1-b). bxl1-b unaffected.
Open follow-up (hypothesis 2, not blocking): bxl1-b on pi-node-004 looks like a migration replica; if bxl1 -> bxl1-b migration is intended, the bxl1 static route should eventually be repointed or removed as part of that migration.